In February 2021, the Empire Center published the first independent analysis of the Cuomo’s administration much-debated directive ordering Covid-positive patients into nursing homes.
The report found that the directive was associated with a statistically significant increase in resident deaths in the homes that admitted the infected patients.
That finding was based on data that the Health Department had belatedly released days earlier, after the Empire Center and the Government Justice Center obtained a court order under the Freedom of Information Law.
The report was co-authored by Bill Hammond, the center’s Senior Fellow for Health Policy, and Ian Kingsbury, a former Research Fellow with the center, and was widely cited by the media and public officials.
Last week, former Cuomo aide Paul Francis questioned the accuracy of Hammond and Kingsbury’s findings as part of a broader defense of Cuomo’s nursing home policies released by the Step Two Policy Project, a think tank founded by Francis. A shorter version of his critique was also published as an op-ed on Sept. 8.
Although it was portrayed as “debunking” the Empire Center’s findings, Francis’ commentary itself was marred by significant factual errors and analytical mistakes. The letter below details those flaws and asks Francis to correct the record.
Dear Paul,
The coronavirus pandemic was a tragic chapter in New York history, especially in its devastating impact on nursing home residents.
Our February 2021 report, “COVID-positive Admissions Were Correlated with Higher Death Rates in New York Nursing Homes,” was a good-faith effort to explore the impact of a hotly debated policy choice on those unfortunate outcomes.
Your recent commentary and op-ed referencing our report makes errors of fact and analysis that led you to inaccurate and unfair conclusions.
As the co-authors, we respectfully request that you correct the record or withdraw your comments.
First and most importantly, you fail to acknowledge that our report included a statewide regression analysis as well as analyses that were limited to the downstate and upstate regions. This may have been a misunderstanding on your part, but it underlies many of your erroneous conclusions.
Our statewide analysis found, as stated in the report, that each admission was associated with an additional .09 fatalities, plus or minus .05.
That statistically significant statewide result was the basis for our conclusion that the March 25 policy was associated with several hundred and possibly more than 1,000 additional deaths.
Second, you falsely state that we “combined the correlated upstate data and the uncorrelated downstate data” in deriving our estimate of the statewide impact.
To the contrary, as explained above, our estimate of the statewide number of deaths associated with the policy was based on our statewide analysis, not a combination of our upstate and downstate analyses.
Third, you mistakenly describe our delineation of downstate and upstate. Our definition of upstate (following regional boundaries used by Empire State Development) excluded the lower Hudson Valley counties of Dutchess, Orange, Putnam, Rockland, Sullivan, Ulster as well as Westchester, New York City and Long Island.
Fourth, you appear to have miscounted admissions in addition to using the wrong counties. You found that “only 27 out of 312 nursing homes upstate … received a total of 509 new admissions.” Our data show that 45 out of 278 nursing homes upstate received a total of 328 new admissions.
Finally, you draw attention to three upstate nursing homes that effectively operated as Covid-only facilities and accounted for a disproportionate share of Covid-related deaths – and suggest that anomalies such as this would render our analysis invalid.
This is a reasonable concern. As made clear in the report, we were working with the limited data available in February 2021, and fully expected that our analysis would need updating as further information became available.
That said, even after removing the three identified Covid-only facilities from the data set, regression analysis of the remaining figures still shows a statistically significant correlation between admissions and deaths – both on a statewide basis and among upstate homes.
In both cases, the correlation would be smaller. Statewide, analysis of the revised data set would indicate that each admission was associated with .07 additional deaths, plus or minus .02. Upstate, each admission was associated .30 additional deaths, plus or minus .09.
The statewide results of this alternative analysis — limited to exclude the three identified Covid-only nursing homes — would suggest the policy was associated with hundreds of additional deaths.
As with our original work, these analyses controlled for the size of each home and the county of its location, and both were statistically significant at a 99 percent confidence level.
We are open to fair-minded and factually accurate critiques of our work, and we readily shared our raw data with you. In writing our report more than three years ago, we had expected that it would quickly be overshadowed by the research of better-qualified public health experts. To our knowledge, that has not yet happened.
We trust you will appreciate our concern about the factual and analytical inaccuracies in your report and understand the importance of setting the record straight. We look forward to your reply.
Sincerely,
Bill Hammond, Senior Fellow for Health Policy, the Empire Center
Ian Kingsbury, Ph.D., Director of Research, Do No Harm, former Research Fellow, the Empire Center